In realistic speech enhancement settings for end-user devices, we often encounter only a few speakers and noise types that tend to reoccur in the specific acoustic environment. We propose a novel personalized speech enhancement method to adapt a compact denoising model to the test-time specificity. Our goal in this test-time adaptation is to utilize no clean speech target of the test speaker, thus fulfilling the requirement for zero-shot learning. To complement the lack of clean speech, we employ the knowledge distillation framework: we distill the more advanced denoising results from an overly large teacher model, and use them as the pseudo target to train the small student model. This zero-shot learning procedure circumvents the process of collecting users' clean speech, a process that users are reluctant to comply due to privacy concerns and technical difficulty of recording clean voice. Experiments on various test-time conditions show that the proposed personalization method can significantly improve the compact models' performance during the test time. Furthermore, since the personalized models outperform larger non-personalized baseline models, we claim that personalization achieves model compression with no loss of denoising performance. As expected, the student models underperform the state-of-the-art teacher models.
more »
« less
Zero-Shot Personalized Speech Enhancement Through Speaker-Informed Model Selection
This paper presents a novel zero-shot learning approach towards personalized speech enhancement through the use of a sparsely active ensemble model. Optimizing speech denoising systems towards a particular test-time speaker can improve performance and reduce run-time complexity. However, test-time model adaptation may be challenging if collecting data from the test-time speaker is not possible. To this end, we propose using an ensemble model wherein each specialist module denoises noisy utterances from a distinct partition of training set speakers. The gating module inexpensively estimates test-time speaker characteristics in the form of an embedding vector and selects the most appropriate specialist module for denoising the test signal. Grouping the training set speakers into non-overlapping semantically similar groups is non-trivial and ill-defined. To do this, we first train a Siamese network using noisy speech pairs to maximize or minimize the similarity of its output vectors depending on whether the utterances derive from the same speaker or not. Next, we perform k-means clustering on the latent space formed by the averaged embedding vectors per training set speaker. In this way, we designate speaker groups and train specialist modules optimized around partitions of the complete training set. Our experiments show that ensemble models made up of low-capacity specialists can outperform high-capacity generalist models with greater efficiency and improved adaptation towards unseen test-time speakers.
more »
« less
- Award ID(s):
- 2046963
- NSF-PAR ID:
- 10318287
- Date Published:
- Journal Name:
- 2021 IEEE Workshop on Applications of Signal Processing to Audio and Acoustics (WASPAA)
- Format(s):
- Medium: X
- Sponsoring Org:
- National Science Foundation
More Like this
-
-
Training personalized speech enhancement models is innately a no-shot learning problem due to privacy constraints and limited access to noise-free speech from the target user. If there is an abundance of unlabeled noisy speech from the test-time user, one may train a personalized speech enhancement model using self-supervised learning. One straightforward approach to model personalization is to use the target speaker’s noisy recordings as pseudo-sources. Then, a pseudo denoising model learns to remove injected training noises and recover the pseudo-sources. However, this approach is volatile as it depends on the quality of the pseudo-sources, which may be too noisy. To remedy this, we propose a data purification step that refines the self-supervised approach. We first train an SNR predictor model to estimate the frame-by-frame SNR of the pseudo- sources. Then, we convert the predictor’s estimates into weights that adjust the pseudo-sources’ frame-by-frame contribution to- wards training the personalized model. We empirically show that the proposed data purification step improves the usability of the speaker-specific noisy data in the context of personalized speech enhancement. Our approach may be seen as privacy-preserving as it does not rely on any clean speech recordings or speaker embeddings.more » « less
-
In this study, we propose to investigate triplet loss for the purpose of an alternative feature representation for ASR. We consider a general non-semantic speech representation, which is trained with a self-supervised criteria based on triplet loss called TRILL, for acoustic modeling to represent the acoustic characteristics of each audio. This strategy is then applied to the CHiME-4 corpus and CRSS-UTDallas Fearless Steps Corpus, with emphasis on the 100-hour challenge corpus which consists of 5 selected NASA Apollo-11 channels. An analysis of the extracted embeddings provides the foundation needed to characterize training utterances into distinct groups based on acoustic distinguishing properties. Moreover, we also demonstrate that triplet-loss based embedding performs better than i-Vector in acoustic modeling, confirming that the triplet loss is more effective than a speaker feature. With additional techniques such as pronunciation and silence probability modeling, plus multi-style training, we achieve a +5.42% and +3.18% relative WER improvement for the development and evaluation sets of the Fearless Steps Corpus. To explore generalization, we further test the same technique on the 1 channel track of CHiME-4 and observe a +11.90% relative WER improvement for real test data.more » « less
-
Direct acoustics-to-word (A2W) systems for end-to-end automatic speech recognition are simpler to train, and more efficient to decode with, than sub-word systems. However, A2W systems can have difficulties at training time when data is limited, and at decoding time when recognizing words outside the training vocabulary. To address these shortcomings, we investigate the use of recently proposed acoustic and acoustically grounded word embedding techniques in A2W systems. The idea is based on treating the final pre-softmax weight matrix of an AWE recognizer as a matrix of word embedding vectors, and using an externally trained set of word embeddings to improve the quality of this matrix. In particular we introduce two ideas: (1) Enforcing similarity at training time between the external embeddings and the recognizer weights, and (2) using the word embeddings at test time for predicting out-of-vocabulary words. Our word embedding model is acoustically grounded, that is it is learned jointly with acoustic embeddings so as to encode the words’ acoustic-phonetic content; and it is parametric, so that it can embed any arbitrary (potentially out-of-vocabulary) sequence of characters. We find that both techniques improve the performance of an A2W recognizer on conversational telephone speech.more » « less
-
Ensemble learning, in its simplest form, entails the training of multiple models with the same training set. In a standard supervised setting, the training set can be viewed as a 'teacher' with an unbounded capacity of interactions with a single group of 'trainee' models. One can then ask the following broad question: How can we train an ensemble if the teacher has a bounded capacity of interactions with the trainees? Towards answering this question we consider how humans learn in peer groups. The problem of how to group individuals in order to maximize outcomes via cooperative learning has been debated for a long time by social scientists and policymakers. More recently, it has attracted research attention from an algorithmic standpoint which led to the design of grouping policies that appear to result in better aggregate learning in experiments with human subjects. Inspired by human peer learning, we hypothesize that using partially trained models as teachers to other less accurate models, i.e.~viewing ensemble learning as a peer process, can provide a solution to our central question. We further hypothesize that grouping policies, that match trainer models with learner models play a significant role in the overall learning outcome of the ensemble. We present a formalization and through extensive experiments with different types of classifiers, we demonstrate that: (i) an ensemble can reach surprising levels of performance with little interaction with the training set (ii) grouping policies definitely have an impact on the ensemble performance, in agreement with previous intuition and observations in human peer learning.more » « less
- Free Publicly Accessible Full Text
-
Accepted Manuscript1.0
- Conference Paper:
- https://doi.org/10.1109/WASPAA52581.2021.9632752
